On Mon Sep 4 00:59:55 2023 +0000, Zebediah Figura wrote:
Oops, sorry about that... Wrt 2/2, my original intent was that these would eventually be exposed as API, and in that case I think byte size is a bit more intuitive to deal with than vec4 size. In that case I intended to fix spirv to treat the size as a byte size, but forgot to do that (and of course the tests didn't catch it, since it just resulted in declaring larger buffers than we actually access).
Makes sense. DXIL contains a byte size so it works well there too.